Digital Screening Platform with Open-Ended Association Questions and Precision Threshold Adjustment

1. A machine learning (ML)-based platform for precision filtering a pool of candidates, said platform comprising computer executable code stored in non-transitory memory that, when run on a processor, is configured to: generate, via an association testing module, an interactive screener web page, wherein, in order to generate the screener web page, the association testing module is configured to: compile and/or access a corpus of data, said corpus comprising a set of words; create and/or access a word association model, said word association model being an ML engine that is trained to derive an association score between a pair of words; and automatically generate an association question that is based at least in part on selecting a base word from the corpus and prompting a question recipient to input a word that is associated with the base word, and wherein the screener web page comprises the association question; transmit the screener web page comprising the association question to a computing device associated with a candidate from the pool of candidates; receive a response that the candidate input to the screener web page in response to the association question, said response comprising the input word; feed, in real time, the base word and the input word to the word association model to derive an association score between the base word and the input word; score the candidate at least in part based on the association score; route the candidate to an interactive on-line survey when the score satisfies a predetermined threshold response score; and route the candidate away from the interactive on-line survey when the score fails to satisfy the predetermined threshold response score; wherein a difficulty level of the screener web page and/or the predetermined threshold response score is determined based on a target level of data quality, said target level of data quality that is associated with accurate completion of the interactive on-line survey.

2. The platform of claim 1 wherein the platform further comprises: a module that tests for a suspicious location of the candidate; a language proficiency module that tests a language proficiency level of the candidate, said language proficiency module comprising a synonym question, a behaviometric analysis, and/or natural language processing (NLP) analysis; and/or an event analysis module that tests for suspicious activity of the candidate.

3. The platform of claim 1 wherein the association testing module is further configured to process the corpus, said processing comprising removing unwanted words, said unwanted words comprising profane words and stop words.

4. The platform of claim 1 wherein the set of words of the corpus is associated with a maximum threshold level of difficulty.

5. The platform of claim 1 wherein the set of words of the corpus is filtered to remove certain predetermined words that are associated with above a threshold incidence of being used by a group of candidates, said group of candidates that are associated with low-quality responses.

6. The platform of claim 1 wherein the set of words of the corpus is filtered to remove words that are within a predetermined association score, as derived by the word association model, to certain predetermined words that are associated with above a threshold incidence of being used by candidates that are associated with low-quality responses.

7. The platform of claim 1 wherein the candidate fails the association question when the association score is below a predetermined threshold score.

8. The platform of claim 7 wherein the predetermined threshold score is 0.25, 0.3, 0.35 or 0.4.

9. The platform of claim 1 wherein the pool of candidates comprises one or more candidates applying to participate in the interactive on-line survey.

10. The platform of claim 1 , wherein the target level of data quality is determined for every survey in a set of interactive on-line surveys.

11. A method for detecting a low-quality input on a digital system, the method comprising: compiling and/or accessing a corpus of words, said corpus of words comprising words that are below a predetermined difficulty level and/or above a predetermined commonality level; selecting a base word from the corpus; generating a screener web page comprising an association question that is based on the base word, said association question prompting a question recipient to input a word that is associated with the base word; transmitting the screener web page to a participant who is accessing the system; receiving an input word, said input word received as input to the screener web page, from the participant, in response to the screener question; feeding, in real time, the base word and the input word to a word association model to derive an association score between said base word and input word, said word association model that is a machine learning (ML) engine that is trained to derive an association score between a pair of words; scoring the participant at least in part based on the association score; routing the participant to an interactive on-line survey when the score satisfies a predetermined threshold response score; and routing the participant away from the interactive on-line survey when the score fails to satisfy the predetermined threshold response score.

12. The method of claim 11 further comprising processing the corpus, said processing comprising removing unwanted words, said unwanted words comprising profane words and stop words.

13. The method of claim 11 wherein the words of the corpus are associated with a maximum threshold level of difficulty.

14. The method of claim 11 further comprising filtering the corpus to remove certain predetermined words that are associated with above a threshold incidence of being used by participants that are associated with low-quality responses.

15. The method of claim 11 further comprising filtering the corpus to remove words that are within a predetermined association score, as derived via the word association model, to certain predetermined words that are associated with above a threshold incidence of being used by a group of participants, said group of participants that are associated with low-quality responses.

16. The method of claim 11 wherein the predetermined threshold score is 0.25, 0.3, 0.35 or 0.4.

17. The method of claim 11 further comprising, when the association score between the base word and input word is below a predetermined threshold association score, flagging the input of the participant as low quality.

18. The method of claim 11 wherein a difficulty level of the screener web page and/or the predetermined threshold response score are set based on a target level of data quality, said target level of data quality that is associated with accurate completion of the interactive on-line survey.

19. A machine learning (ML)-based digital filtering system for providing quality data for a survey while maintaining a level of participant diversity, said system comprising computer executable code stored in non-transitory memory and run on a processor, said system further comprising: a screener module that tests a data quality level; and an anti-fraud module that tests for fraudulent activity; wherein: the screener module is configured to generate a pre-survey screen web page that is transmitted to a computing device associated with a participant, wherein the pre-survey screen web page comprises an association question that tests an association ability level of the participant, and the screener module is configured to: compile and/or access a corpus of data, said corpus comprising a set of words; create and/or access a word association model, said word association model being an ML engine that is trained to derive an association score between a pair of words; and automatically generate the association question that is based at least in part on selecting a base word from the corpus and prompting the participant to input a word that is associated with the base word; the system is configured to receive a response word that is input by the participant to the pre-survey screen web page; the system is configured to score the response word, said score being based at least in part on the association score derived by the word association model between the base word and the response word; and when the score fails to achieve a predetermined threshold response score, the system is configured to flag the participant and/or reject the participant from the survey; and wherein a difficulty level of the pre-survey screen web page and/or the predetermined threshold response score is determined based on a target level of data quality.

20. The system of claim 19 further comprising: a module that tests for a suspicious location of the participant; a language proficiency module that tests a language proficiency level of the participant, said language proficiency module comprising a synonym question, a behaviometric analysis, and/or natural language processing (NLP) analysis; and an event analysis module that tests for suspicious activity of the participant.